Sivers Semiconductors (SIVEF) falls 39% in four sessions as wireless losses eclipse photonics

The number that surprised me

Two notes on reading that table, because the numbers look bigger than the ones in the headlines. The three EBITDA lines do add up: negative SEK 38.4 million, negative SEK 28.3 million and negative SEK 31.6 million of unallocated central costs give the group's negative SEK 98.3 million. The figure Sivers leads with is adjusted EBITDA of negative SEK 35.5 million, and the gap between the two is SEK 62.7 million of add-backs, mostly SEK 50.3 million of share-based payment expense plus SEK 12.4 million of US listing preparation costs. Most of that share-based charge is the Swedish social tax provision we wrote about when the report landed, and it lands in the group-wide line rather than in either segment.

Even after all that, wireless is the segment that deteriorated fastest, and it's the bigger one: SEK 208.5 million of 2025's SEK 302.8 million of net sales against SEK 94.3 million for photonics. It's also where the orders are. The $8.2 million production order from ALL.SPACE for Ka-band beamforming chips and the initial $3 million from Tachyon Networks are both real and both signed, and both sit in the business that lost SEK 38.4 million.

The wireless order book is still worth having. But it's hard to call near-term wireless revenue the disciplined choice and photonics the speculative one when photonics lost less money last quarter. Right now neither pays for itself.

The part most people have forgotten

In August 2024 Sivers signed a letter of intent to merge Sivers Photonics into byNordic Acquisition Corporation, a listed SPAC, which would have made photonics its own public company with Sivers keeping the majority. On November 11 the board put the discussions on hold and released byNordic to find someone else, blaming conditions in the US market for small caps and SPACs.

Chairman Bami Bastani, in that release: "While we agree with ByNordic's thesis that our photonics business is highly undervalued in Sivers' current market capitalization with its critical positioning in the upcoming AI datacenters, the capital markets for successfully executing SPAC mergers remain challenging."

Photonics is the main reason this company got international attention at all, and it's most of why the stock is still up about 450% over the past twelve months, even after the week it just had. Two years ago the board's answer to that business being undervalued inside Sivers was to move it out of the building. Worth remembering now that the argument has flipped to whether photonics deserves the focus at all.

Who is left watching

Wednesday morning DNB Carnegie ended its commissioned research coverage, and the analyst wouldn't say why. That was paid coverage, funded by Sivers itself, and its last fair value range was SEK 3.0 to SEK 7.5. The only other public target is Redeye's SEK 5.8 from October 2025. So when Placera ranked Sivers first on the Swedish market for analyst downside at 68% on Wednesday, understand what's underneath it. Every published target still sits far below SEK 22.10, so the downside those numbers imply is substantial and it's pointing one way. But two targets, one ten months old and one just withdrawn, isn't a consensus, and it shouldn't be read as one.

Short sellers came back while this happened. Sweden's Finansinspektionen puts aggregate net short interest at 4.4% of the share capital, roughly 15.7 million shares, up from 3.13% in early August when we last went through the register. Sweden only publishes a holder's name above 0.5%.

Those three account for 2.02 of the 4.4 points. The rest belongs to holders between 0.1% and 0.5%, who never get named. D.E. Shaw has been building since August 3, when it first showed up at 0.62%. And one name there isn't a fund with a view: Citadel Securities is a market maker, so its short line is usually inventory from making markets rather than a bet.

Where I land on the listing

Preparations for the New York dual listing are now expected to finish in the first half of 2027, with the decision depending on market conditions. Plenty of people read that as another slip. I'd read it the other way. Ready and listed aren't the same thing, and Sivers gets one first impression with a much larger US investor base. Doing it before the product ramps are visible spends that for nothing. The Jabil module hits beta builds in Q4 2026 with initial production orders expected in the first half of 2027, and chief executive Vickram Vathulya puts the long-term financial model at "2028 onwards." Listing into that gap, with a restatement and an unanswered short report still behind you, is how you get a bad first print.

What I'd want is the company saying that out loud. Pick the checkpoint, name it, and let people hold you to it. The one thing the report and the call didn't give anybody was a date somebody's job depends on.

For the record, I don't own SIVEF and never have. That's luck more than judgment. I've watched the photonics names all year and this is the one I could never get comfortable enough with to size.

Third-quarter numbers land November 26.